5D3 Raw post-processing options - Pros and Cons

How does Resolve 10 DeBayer compare to ACR? R9 is a lot softer and no NR (free version) makes it not very useful compared to ACR (one time transcode to a good codec is the only time issue).
 
Can the frame format be changed to something more suitable for anamorphic? such as square or 4:3 HD, rather than 16:9 HD?
 
IMO, there's no contest between AE and R10... at all. Jump on ebay or go over to BMCuser to see if you can pick up a cheap Dongle second hand from someone selling it.

The CinemaDNG debayer settings for motion picture/video have been updated, and having been testing R10 for the past few weeks, sorry... I just can't see how or why anyone would want to finish anything outside of R10, especially RAW.

I consider RED's software to be very solid for a free, specialized application, but it can't hang with Resolve's Debayer, Color, etc.
 
I haven't done a lot of testing, but to me it seems R9 is just as sharp as ACR: http://www.similaar.com/foto/blackmagic-workflow/index.html

Comparing R9 to ACR (default sharpness 25) in your example- on my iPad 4 (latest gen) the ACR image is not only clearly sharper, but has better color. When comparing a variety of organic shots with 5D3 RAW, ACR looked even better vs R9. I'm not sure if ACR is doing this, but state of the art de-Bayer methods do NR and other processing at the same time as the de-Bayer- improving reconstruction quality (more accurate- not just more detailed). 5D3 is only 1.9K Bayer- needs all the help it can get.

Regarding buying a used dongle- would wait for the Pocket camera to be available. I'll look into borrowing a dongle from a friend.
 
R9/10 with NR and high-quality de-Bayer (matching ACR) appears to require the full version (does full version have de-Bayer options?). AE+ACR to DNxHD or ProRes is fairly fast on a 12-Core MacPro resulting in smaller files (can delete intermediate files), edits in real-time and quality is excellent. That said, if R10 can match ACR de-Bayer & NR, would be worth using if it saved time. Again, unlike higher res cameras, the 5D3 really needs every ounce of de-Bayer quality possible (1.9K source Bayer RAW).
 
*question goes unanswered... slowly and quietly side steps towards the door, and awkwardly leaves the room. =)
 
Can the frame format be changed to something more suitable for anamorphic? such as square or 4:3 HD, rather than 16:9 HD?

Yes- 4:3 is supported in frame format (and many more) as well as Anamorphic mode for Live View (to unsqueeze image while shooting). Haven't shot anamorphic yet; just looked up settings and tried in Live View.
 
Kholi. Is R10 stable? Has it crashed on you?

R10 lite is free btw. No dongle needed and supports up to 4k!

Actually, R10 is as stable as R9 on my MBPR, which I suspect's due to the focus on new architecture and support for Mac going forward. I've had LESS crashes in R10 working with footage than I have in R9, but knock on wood because I haven't messed with as much RAW data in it yet. We've finished some small commercials in the beta over the last month+ (we were beta testers), so it was fine.

I imagine that this iteration will improve faster.

The reason I mentioned the dongle's because that's where you get the Noise Reduction and a few other locked options, but certainly, you can use R10 Lite to get the hang of the application until then.

I'm still not sold on ACR as the better Debayer. R10 processes CinemaDNG with new sharpness settings as well, which is quite useful. But, I suppose maybe I should look harder.
 
The biggest difference between Resolve and ACR is the speed of the workflow. Resolve will beat ACR hands down every time. Working with a proxy based workflow makes Resolve just that much better. Those that are trying to edit DNG natively, try it on a 30 minute project and not 2 or 3 minutes of footage. You can load all your RAW footage into Resolve, make some quick proxies, edit, then finish in Resolve. Now getting the best image is still up for debate. But I feel the people that are saying ACR is better are just used to ACR and not very proficient in Resolve.

R10 is looking very good and is running quite smooth on my MBPr as well.
 
Yes- 4:3 is supported in frame format (and many more) as well as Anamorphic mode for Live View (to unsqueeze image while shooting). Haven't shot anamorphic yet; just looked up settings and tried in Live View.
Thank you much. I was hoping so. Hopefully the resolution in 4:3 can be more dense, so as to not lose resolution in the same way 16:9 HD would after a crop. Thanks again.
 
Thank you much. I was hoping so. Hopefully the resolution in 4:3 can be more dense, so as to not lose resolution in the same way 16:9 HD would after a crop. Thanks again.

I was able to 'activate' 3:2, which is 1920x1280. 4:3 would be 1920x1440- requiring a very fast card. It's probably possible to crop down a bit to lower CF card requirements, but didn't explore any further. The folks over here know a lot more: http://www.magiclantern.fm/forum/index.php?topic=6215.0
 
The biggest difference between Resolve and ACR is the speed of the workflow. Resolve will beat ACR hands down every time. Working with a proxy based workflow makes Resolve just that much better. Those that are trying to edit DNG natively, try it on a 30 minute project and not 2 or 3 minutes of footage. You can load all your RAW footage into Resolve, make some quick proxies, edit, then finish in Resolve. Now getting the best image is still up for debate. But I feel the people that are saying ACR is better are just used to ACR and not very proficient in Resolve.

R10 is looking very good and is running quite smooth on my MBPr as well.

IIRC the 12-Core MacPro with AE+ACR CS6 ran 1/4 real-time converting to DNxHD 10-bit 172Mbit/s footage. That's 2 hours to convert 30 minutes. Once done, footage can be edited anywhere in real-time, and the files wouldn't fill up my MBP SSD (no need to carry an external USB3 drive). If something faster needed time-wise, I'd use my FS700 (internal codec or high-bitrate 422 codec and Nanoflash); you'd use your C100.

While Resolve is very complex, it's not that hard to figure out how to use, especially with google and youtube available. That said, R9 Lite was noticeably lower quality vs. AE+ACR due to weaker de-Bayer method and no available noise reduction. Again, the latest advanced de-Bayer methods include de-noising during de-Bayer (all part of "reconstruction"), so even with a separate de-noise pass in Resolve (or elsewhere, such as with Neat Video), no example yet showing better results vs. ACR which performs de-noise and de-Bayer at the same time (and possibly using one of the latest advanced techniques combining both algorithms at once during reconstruction). The very latest algorithms use multiple frames to de-noise and de-bayer (super-resolution, etc.).
 
IIRC the 12-Core MacPro with AE+ACR CS6 ran 1/4 real-time converting to DNxHD 10-bit 172Mbit/s footage. That's 2 hours to convert 30 minutes.

A 30 minute end project probably has atleast 2-4 hours of footage.

The best thing about ACR is it's excellent sharpening algorithm, but that's just that. Sharpening. At 0 the image looks natural. ACR also has excellent highlight recovery but it does do some tricks to get there. You can also use ACR's sharpening and highlight recovery for jpeg files in photoshop and be surprised by how well it works in stuff that's not RAW. I'm not sure Resolve R10 has weaker debayering, I just think you are confusing less sharpening with "weaker".

I consider de-noising and sharpening completely different things than debayering. I thought the image should be "as is" without software tricks? You can do all kinds off tricks to FS700 material too, a bit of denoise with neatvideo, then a bit of grain and some highlight roll-off with a curve. And suddenly AVCHD also takes time to process.
 
Using ACR to transcode into something more manageable is not optimal in terms of workflow: you either have to relink back to the original RAW clips later on, or you have to color correct every single RAW clip before you transcode them.
If the AE/Premiere update brings real-time DNG playback, it's a game changer: I wouldn't have to learn Resolve (which I tried, and found puzzling), and I would be able to work just as I do with standard H.264 files (edit first, color correct only the chosen bits).
 
A 30 minute end project probably has atleast 2-4 hours of footage.

The best thing about ACR is it's excellent sharpening algorithm, but that's just that. Sharpening. At 0 the image looks natural. ACR also has excellent highlight recovery but it does do some tricks to get there. You can also use ACR's sharpening and highlight recovery for jpeg files in photoshop and be surprised by how well it works in stuff that's not RAW. I'm not sure Resolve R10 has weaker debayering, I just think you are confusing less sharpening with "weaker".

I consider de-noising and sharpening completely different things than debayering. I thought the image should be "as is" without software tricks? You can do all kinds off tricks to FS700 material too, a bit of denoise with neatvideo, then a bit of grain and some highlight roll-off with a curve. And suddenly AVCHD also takes time to process.

De-Bayering, de-noising, and sharpening/detail-enhancement are all in the same class of reconstruction algorithms. Higher quality algorithms produce less artifacts. I was not able to make R9 match ACR for a high detail low artifact result. Unmask or convolution sharpening won't cut it. Perhaps Resolve can someday leapfrog ACR with temporal super-resolution de-Bayering, denoising, sharpening, and up-rezzing.
 
Using ACR to transcode into something more manageable is not optimal in terms of workflow: you either have to relink back to the original RAW clips later on, or you have to color correct every single RAW clip before you transcode them.
If the AE/Premiere update brings real-time DNG playback, it's a game changer: I wouldn't have to learn Resolve (which I tried, and found puzzling), and I would be able to work just as I do with standard H.264 files (edit first, color correct only the chosen bits).

Of course it depends on the length of what you are creating, for me AE+ACR produces great quality for little time and effort. 10-bit is plenty for my web targeted content. ProRes supports 12-bit, so no loss in bit depth after de-Bayer and plenty of color info to color correct. I'd even pay the full $1k for Resolve (since Pocket cam isn't out) if it produced higher quality vs ACR... (Check tutorials on YouTube- it's not a great CogSci-based GUI, however after tutorials isn't hard).
 
De-Bayering, de-noising, and sharpening/detail-enhancement are all in the same class of reconstruction algorithms. Higher quality algorithms produce less artifacts. I was not able to make R9 match ACR for a high detail low artifact result. Unmask or convolution sharpening won't cut it. Perhaps Resolve can someday leapfrog ACR with temporal super-resolution de-Bayering, denoising, sharpening, and up-rezzing.

ACR doesn't do denoising with multiple frames, I mean it is designed for stills. It takes one image at a time so why do you keep mentioning temporal stuff?
 
ACR doesn't do denoising with multiple frames, I mean it is designed for stills. It takes one image at a time so why do you keep mentioning temporal stuff?

Perhaps re-read my posts. There's no suggestion that ACR does anything with multiple frames (temporal, as Neat Video and other tools can). ACR's single frame, single step de-Bayer, de-noise, and sharpen out-performs Resolve currently. Some folks mentioned the patent- perhaps that's an issue for Resolve.

I have mentioned that the most advanced papers on de-Bayer are using temporal methods, which also include the benefit of de-noising and go beyond just sharpening: actual, real resolution enhancement. Super resolution removes aliasing and provides real resolution increases (vs. just sharpening). Lots of papers for those curious about the algorithms and math: https://www.google.com/search?q=tem...solution&oq=temporal+debayer+super+resolution

In that last post I suggested that BM might be able to go beyond what ACR is doing by using one of these latest temporal methods (or a novel one of their own).
 
Back
Top